lable at ScienceDirect

Journal of Loss Prevention in the Process Industries 23 (2010) 253–259

Contents lists avai

Journal of Loss Prevention in the Process Industries

journal homepage: www.elsevier .com/locate/ j lp

Explosions in closed pipes containing baffles and 90 degree bends

Robert Blanchard*, Detlef Arndt, Rainer Gratz, Marco Poli, Swen ScheiderBAM Federal Institute for Materials Research and Testing, Chemical Safety Engineering, Explosion Protection and Risk Assessment, Unter den Eichen 87, 12205 Berlin, Germany

a r t i c l e i n f o

Article history:Received 11 May 2009Received in revised form3 September 2009Accepted 3 September 2009

Keywords:Explosion enhancementDeflagration to detonation transitionBendBaffle

* Corresponding author. Tel.: þ49 30 8104 4258; faE-mail address: robblanchard@hotmail.com (R. Bla

0950-4230/$ – see front matter � 2009 Published bydoi:10.1016/j.jlp.2009.09.004

a b s t r a c t

There is a general lack of information on the effects of full-bore obstacles on combustion in the literature,these obstacles are prevalent in many applications and knowledge of their effects on phenomenaincluding burning rate, flame acceleration and DDT is important for the correct placing of explosionsafety devices such as flame arresters and venting devices. In this work methane, propane, ethylene andhydrogen–air explosions were investigated in an 18 m long DN150 closed pipe with a 90 degree bend andvarious baffle obstacles placed at a short distance from the ignition source. After carrying out multipleexperiments with the same configuration it was found that a relatively large variance existed in themeasured flame speeds and overpressures, this was attributed to a stochastic element in how flamesevolved and also how they caused and interacted with turbulence to produce flame acceleration. This ledto several experiments being carried out for one configuration in order to obtain a meaningful average.It was shown that a 90 degree bend in a long tube had the ability to enhance flame speeds and over-pressures, and shorten the run-up distance to DDT to a varying degree for a number of gases. In terms ofthe qualitative effects on these parameters they were comparable to baffle type obstacles with a blockageratios of between 10 and 20%.

� 2009 Published by Elsevier Ltd.

1. Introduction

Explosions in the process industry are still a significant problemleading to injuries, death, destruction of equipment and downtime.As a consequence there is a need in many chemical processes forprotection against propagation of unwanted combustion phenomenasuch as deflagrations and detonations (including decompositionflames) in process equipment, piping and vent manifold systems(vapour collection systems) (Grossel, 2002). Environmental andsafety concerns relating to volatile organic solvents (VOCs) have also,in recent years, lead to a vast increase in the amount of piping andnumber of systems handling flammable fuel air mixtures for collec-tion, containment and transport. Control over these mixtures is alsorequired until flammable gases can be destroyed or safely discharged(Newsholm, 2004). In many of the above situations, in order to aidATEX compliance, correctly placed and specified flame arresters areneeded, dependent on the conditions they are likely to encounter.However, there is still some uncertainty over where best to locatethese devices and concerns have been raised with safety standardsfor flame arrestors in regards to the lack of knowledge of wheredeflagration to denotation will/can occur in a pipe and what factorscan contribute to this effect (Oakley & Thomas, 2004). These concerns

x: þ49 30 8104 1217.nchard).

Elsevier Ltd.

persist still to this day and as part of a safety concept related to flamearrestor positioning it is important to be able to predict the mode ofburning at various points in a pipe.

Explosions in pipes and ducts, flame acceleration and the tran-sition from deflagration to detonation are well researched subjects(Ciccarelli & Dorofeev, 2008). However, research in this area tendsto concentrate on the effects of baffle type obstacles or items in thepath of the flow (Ibrahim & Masri, 2001; Masri, Ibrahim, Nehzat, &Green, 2000). Tube bends, for example, are full-bore obstaclesused extensively in industrial applications, however little is knownabout their effects on flame acceleration, overpressure enhance-ment and their contribution to DDT.

1.1. Research aim

The aim of the research presented in this paper is to communicatedata from explosion experiments in an industrial-scale closed pipecontaining a 90 degree bend in order to better understand andquantify the affects these types of obstacle can have on flame accel-eration and DDT. Experimental data is also compared to that foundusing baffle type obstacles in the same position as the 90 degree bend.

1.2. Background

Fluid and particle flow through pipe bends is a well understoodsubject due to its associated practical problems in the process

R. Blanchard et al. / Journal of Loss Prevention in the Process Industries 23 (2010) 253–259254

industry (Berger, Talbot, & Yao, 1983; Burgess, 1971; Green, 1999),however little research has been carried out on explosions throughpipe bends, a complicated problem involving the interactionbetween fluid dynamics, heat transfer and (turbulent) combustion.In the first work carried out on this theme Phylaktou, Foley, andAndrews (1993) showed that with a short tube a 90 degree bendcan enhance to both the flame speed and the overpressure formethane–air explosions compared to similar experiments carriedout in straight pipes. The flame speed in these experiments wasenhanced by a factor of approximately five and was equated to theeffects of a baffle with a blockage ratio of 20% at the same position.

Sato, Sakai, and Chiga (1996) investigated the effects of ignitionposition on the shape of the flame front and the flame speed formethane–air explosions using an open ended small square channelcontaining a 90 degree bend. However, only a limited number ofexperiments were carried out and no comparison was given to anexperimental set-up without the bend.

A 24% enhancement of the flame speed after a 90 degree bendplaced half-way down a tube was observed in propane–air exper-iments by Chatrathi (1992), the pipe diameter used for theseexperiments was 152.4 mm and the pipe was open at the endfurthest from the ignition source.

Observations of the flame front when travelling though a rect-angular 90 degree bend were made by Zhou, Sobiesiak and Quan(2006), who showed that after initially propagating as a flat flamethe flame front takes on the tulip configuration (Clanet & Searby,1996; Gonzalez, Borghi, & Saouab, 1992). As the flame reached thebend the upper tongue (the one propagating towards the outside ofthe bend) of the tulip slowed down, relatively. Whereas the lowertongue the one propagating towards the inside of the bend beganpropagating more quickly around the inside of the bend, an effectnamed ‘‘flame shedding’’ by the authors. 3-D particle modelling ofthe flow around the bend showed that large vortexes were createdjust downstream of the inside wall of the bend while flow followeda more streamlined pattern around the outside of the bend. Thiswas in good agreement with latter constant temperatureanemometry (CTA) observations made by Lohrer, Hahn, Arndt, andGratz (2008) who showed that a bend induced a significantincrease in turbulence over the first 30% of the inner diameter ofthe pipe immediately after the bend. Whereas, only a relativelysmall amount of turbulence was induced around the outer side.Unfortunately in the study of Zhou et al. the bend was relativelyclose to the end of the tube giving little opportunity to observe theeffect of the bend on downstream flow patterns.

Explosions (propane–air) have also been carried out in coiledpipes and pipes with multiple U-shaped bends (Frolov, 2008),while these configurations were able to produce fast DDTs, theyare generally of more interest for pulse detonation engines

Fig. 1. Schematic of the pipe use

(Roy, Frolov, Borisov, & Netzer, 2004), where fast DDT isa requirement, rather than for industrial-scale pipework carryingpotentially flammable gas.

2. Experimental

2.1. Apparatus

A horizontal DN150 steel pipe, shown in Fig. 1, was used for thefollowing tests (d¼ 159 mm). This pipe was made up of a number ofsegments ranging from 2 to 5 m in length, bolted together witha gasket seal in-between the connections and blind flanges at bothends. Evacuation prior to introduction of the test gas confirmed nosignificant leaks were present in the pipe. Early work employingconstant temperature anemometry techniques (Lohrer, Drame,Schalau, & Gratz, 2008) had also shown that these connectionsintroduced no significant turbulence inducing element to flowalong the pipe. The total length of the pipe was approximately 18 m(L), giving an L/d ratio of 112.

Several tests utilised a 90 degree bend segment, this bend hada radius of 0.285 m and added a further 0.447 m to the length of thepipe (based to the centreline length of the segment). Initially this 90degree segment was placed at a distance of 2 m from the ignitionsource. Due to the constraints of the dimensions of the room wheretests were carried out this was the maximum allowable distancebefore the pipe bend. Several tests also utilised baffles, these weremade from 3 mm thick sheet metal and were placed in-betweenconnections in the pipe. Baffles used in these tests had sharp 90degree corners and were made in 10% blockage ratio increments.The blockage ratio is defined as the surface area of the obstacle inthe pipe over the cross sectional area of the pipe.

2.2. Sensors and data collection

The pipe contained sensor penetrations every 0.5 m in the axialdirection which were used to position photodiodes and pressuretransducers. Penetrations not used were filled with plugs which satflush with the inside of the tube.

The position, and hence the speed, of the flame front wasdetermined using BPY 62 Silicon NPN phototransistors (OSRAM)placed at metre intervals along the entire length of the pipe.

The pressure at various points along the length of the pipe wasrecorded using piezoelectric pressure transducers (PCB M113A22,1,5 � 0.005 mV/kPa) with the signal being processed by a PCB 481Sensor Signal Conditioner. Pressures reported are the absolutemaximum side-on overpressures recorded during experiments andthe maximum side-on overpressures after a smoothing functionhad been applied to original pressure-time signals.

d for explosion experiments.

R. Blanchard et al. / Journal of Loss Prevention in the Process Industries 23 (2010) 253–259 255

All sensors were flush with the inside surface of the tube inorder to avoid any additional turbulence enhancement. Data fromall sensors was collected at a frequency of 0.18 MHz.

2.3. Gas mixtures

Gas mixtures were produced using the partial pressure methodand mixed by a paddle in a rocking tube. After sufficient timethe gas mixture was introduced to the evacuated pipe, to thedesired pressure. All gas mixtures used during these experimentswere at stoichiometric concentration and ambient pressure, unlessotherwise stated. The gases used for the following experimentswere methane, propane, ethylene and hydrogen.

Gas mixtures were ignited at one end of the pipe using a meltingwire (ignition energy approximately 10 J). The position of the igni-tion source has a significant effect over the initial propagation of theflame and the resulting flame speeds and overpressures (Phylaktou& Andrews, 1991a, 1991b), so was kept at a constant position flush atthe centre of one of the blind flanges. Ignition was detected usinga phototransistor placed close to the ignition source, this photo-transistor also triggered the data logging system.

For most experimental configurations a number of tests (minimumthree) were carried out depending on the reproducibility of the flamespeeds and overpressures.

3. Results and discussion

3.1. Methane experiments

Experiments were first carried out with methane–air mixturesin a straight pipe. These initial experiments were then compared toa series of methane–air explosions carried out utilising the 90degree bend and baffle obstacles detailed in the previous section.The first striking observation was the variability of flame speedsand overpressures measured during experiments using the sameapparatus, test gas mixture and method.

The maximum flame speed along the length of the pipe for sixmethane–air explosions varied from approximately 20–80 ms�1,many of these explosions also showed two relatively large peaks in

Fig. 2. Flame speeds for methane–air explo

the flame speed between 30 and 60% of the pipe length (all flamespeed data for these and other experiments discussed in this papercan be found online at http://www.XXXlink to Elsevier websi-teXXX). It is difficult to determine the exact cause for the variancein these resulting flame speeds as there are complicated interac-tions involved, and a stochastic element in how the flame evolvesand then causes and interacts with turbulence to produce flameacceleration. All pre-set variables were controlled within theconfines of experimental error. However, possible sources of thisvariance could have derived from the exact nature of the ignitionevent, localised wall characteristics and small changes in the gasconsistency etc. This phenomena should be further investigated asit could have significant consequences for the standard testing ofitems such as flame arresters (European Standard EN 12874, 2001)where, for example, a pre-set number of deflagrations should beunable to pass through an arrester element in order for it to becertified. These tests are assumed to be on the conservative-sidedue to the tube being closed at both ends. However, if explosionswere all at the lower end of the spectrum in terms of flame speedsand overpressures then the flame arrester may not be able towithstand harsher conditions produced by the same gas mixture insimilar equipment at another time. The oscillations seen in theflame speed are a common occurrence during these types ofexplosion and are attributed to complicated interactions betweenflame acceleration, quenching on the walls of the pipe, changes inthe shape of the flame front (Phylaktou & Andrews, 1991a, 1991b)and interactions with pressure waves reflected by the closed end ofthe pipe. The pressure measured at a number of points along thelength of the pipe was seen to be relatively similar regardless of themeasurement position. For these ‘‘slow deflagration’’ experimentsthis was due to the pressure waves caused by the explosiontravelling considerably faster than the flame front, meaning thata pressure sensor in these experiments saw approximately thesame pressure profile wherever it was placed in the pipe.

Examples of flame speeds for methane–air explosions in fourpipe configurations; straight, and with a 10, 20 and 30% BR baffle areshown in Fig. 2, the vertical line in the graph represents the positionof the obstacle. It should be noted that, in this Figure only oneexample of each experimental configuration is displayed in order to

sions with a 10, 20 and 30% BR baffle.

R. Blanchard et al. / Journal of Loss Prevention in the Process Industries 23 (2010) 253–259256

illustrate an effect. Also, in the following discussion flame speeds forthese and other explosions are always compared to those found inthe straight configuration, which is taken to be the base case.

The use of all three baffles resulted in higher downstream flamespeeds being recorded as the blockage ratio of the baffles increased.This effect has been observed by many research groups and iswidely accepted to be due to turbulence inducing elementsenhancing the heat and mass transfer in reactive flows (Andrews,Herath, & Phylaktou, 1990; Phylaktou & Andrews, 1991a, 1991b;Starke & Roth, 1989).

No large difference in the initial flame speeds for the straightconfiguration and the configuration with the 10% BR baffle wereseen, indicating that the 10% BR baffle has little effect on upstreamflow. The 20 and 30% BR baffles induced some flame accelerationbefore the baffle possibly due to flow contraction. However, themajority took place behind the obstacle. Interestingly, the 10%baffle configuration shows some acceleration after the obstacle butthe maximum flame speed is recorded further down the pipe afterthe flame had returned to similar speeds to that seen in the straightpipe configuration. Conventional wisdom would suggest that theaccelerating effect of the baffle should be seen directly downstreamof the obstacle. Therefore, to better explain these effects moresophisticated measurement techniques, visualisation of the flameand CFD would be needed.

The effect of a 90 degree bend on methane–air explosions at thesame position as the aforementioned baffles was also compared.Examples of flame speeds produced using this, the straight and 10%BR baffle configurations are shown in Fig. 3. The bend configurationdisplayed slightly higher flame speeds upstream of the obstacleposition than the straight and 10% BR baffle configurations. Bothobstacles produced similar maximum flame speeds (approximately70 ms�1), however, the maximum flame speed seen during theexplosion in the bend configuration occurred immediately after theobstacle, as would be expected from the accepted theory of flameacceleration by an obstacle. For the baffle configuration a smalleracceleration occurred immediately after the obstacle followedby a deceleration before a final acceleration during which themaximum flame speed was recorded. It is acknowledged that theform of the flame front will be different as it approaches and passesthese obstacles, and therefore there will be differences in the waywhich the flame interacts with the turbulence caused by theobstacle, making a true quantitative comparison between the twois very difficult. However, knowledge of the qualitative effects of

Fig. 3. Flame speed for methane–air explosions in straight, 10% BR baffle and 90 degreebend configurations.

these obstacles can be important in predicting flame accelerationphenomena in pipes.

The methane explosions are displayed in terms of theirmaximum flame speeds, overpressures (smoothed and peak) andmaximum rates of pressure rise in Table 1. The maximum smoothedoverpressure is the maximum recorded pressure after a smoothingfunction had been applied to the pressure-time history, whereas themaximum peak overpressure was the actual highest overpressurerecorded by the sensor with no smoothing function applied.

Although there is some scatter, general trends in maximumflame speed can be seen from these experiments. Firstly there wasno large difference in the maximum flame speeds for methane–airexplosions when comparing the straight and bend configurations,indicating that the bend in this position induces only a relativelysmall amount of turbulence with which the flame can interactand accelerate. However, the maximum flame speed for the bendconfiguration was generally found at around 20% of the pipe length,directly after the obstacle position (w10%), whereas the maximumflame speeds for the straight configuration was found between 30and 50% of the pipe length. This would indicate that, as with the20 and 30% BR baffle configurations, the turbulence induced bythe bend is significant and does cause an initial acceleration of theflame. Further downstream, may be due to quenching on the wallsof the pipe or changes in the flame front geometry, the flame speeddecelerated to speeds below those recorded in the straight pipeconfiguration between 30 and 50% of the pipe length.

The maximum flame speeds for 10% baffle configuration werealso seen between 30 and 50% of the length of the pipe, and were onaverage higher than those produced in the straight pipe configura-tion. This suggests that while the baffle produced no immediateeffect of increasing the rate of burning directly behind the obstacle itcontributed to whatever effects were acting to accelerate the flamespeed later on in the straight pipe configuration. As understandingthe exact nature of these effects would require the quantification ofcomplicated interactions between several factors (fluid dynamics,heat transfer and (turbulent) combustion), it is beyond the scone ofthis work and would require visualisation of the flame and moresophisticated measurement techniques. However, these resultsprovide extremely useful information with regard to flame arresterplacement and determining likely modes and rates of burning ina given system.

Maximum overpressures, both smoothed and peak, and the rateof pressure rise for each of these experimental configurationsshowed a large variation. For example, the maximum smoothedoverpressure for experiments with the straight pipe configurationwere recorded between 1 and 1.8 bar. This variance is again animportant factor to take into account when determining how manyexplosions should be conducted during a standard test procedure.

Representative pressure vs. time profiles for the various config-urations are shown in Fig. 4. The rate of pressure rise seen in theseplots supports observations made concerning flame speeds. Thestraight, bend and 10% BR baffle configurations displayed relatively

Table 1Flame speed and overpressure characteristics of methane explosions.

Configuration Max. flamespeed/ms�1

Max.overpressure(Smoothed)/bar

Max.overpressure(Peak)/bar

Max. rateof pressurerise/bar s�1

Average SD Average SD Average SD Average SD

Straight 48.1 17.2 1.5 0.2 2.0 0.3 4.2 1.0Bend 51.1 8.9 1.5 0.2 2.1 0.4 4.6 1.510% BR Baffle 82.2 20.9 1.4 0.2 2.4 0.5 5.0 2.520% BR Baffle 100.3 2.6 1.0 0.2 1.4 0.3 6.0 0.330% BR Baffle 149.8 20.3 1.1 0.1 1.6 0.3 6.8 1.6

Fig. 4. Overpressure vs. time for methane–air explosions in straight, 10, 20, 30% BRbaffle and 90 degree bend configurations.

Fig. 6. Overpressure vs. time for propane–air explosions in straight, 20% BR baffle and90 degree bend configurations.

R. Blanchard et al. / Journal of Loss Prevention in the Process Industries 23 (2010) 253–259 257

similar flame speeds and rates of pressure rise, while the 20 and 30%baffle configurations showed higher rates of pressure rise whichwere associated with the observed higher burning speeds.

3.2. Propane experiments

Fig. 5 shows representative results from propane–air explosionsutilising the straight pipe configuration and configurations witha 90 degree bend and a 20% BR baffle at 2 m from the ignition source.

Experiments carried out by Phylaktou et al. (1993) showedsimilar effects on the maximum flame speed when using a 90degree bend or a 20% BR baffle. In their experiments the flamespeed was enhanced by a factor of up to 6 times depending on theobstacle position. An enhancement factor of 2 was observed duringthese experiments which was due to the bend being placed closerto the ignition position giving the flame less time to acceleratebefore the obstacle and hence produce less turbulence downstreamof the bend. In the experiments carried out by Phylaktou et al. it wasassumed that the position which gave the maximum flame speedsfor the baffle obstacle would also give the maximum flame speedswhen using the bend. This may not always be the case, therefore,further experiments using different bend positions need to becarried out, however, unfortunately due to the dimensions of thetest chamber this was not possible at present.

Fig. 5. Flame speeds for propane–air explosions.

Pressure vs. time records for propane experiments are shownin Fig. 6, from these it can also be seen that the burning ratewas enhanced by an equal amount for both the bend and baffleconfigurations.

The average maximum flame speeds recorded during propane–air explosions are presented in Table 2 along with data on themaximum overpressures and maximum rates of pressure rise.While there is some spread in the results generally speaking theflame speeds produced using the bend and 20% BR configurationslay within the same range, reinforcing the hypothesis that a 20% BRbaffle enhances the burning rate to the same degree as a 90 degreebend.

The maximum overpressures and rates of pressure rise producedduring the experiments utilising the baffle and bend showed slightlyhigher values compared to the straight configuration, this may beexpected due to the higher flame speeds recorded during theseexperiments but there is too much spread in the data to draw anyfirm conclusions.

One suspected DDT event was observed during the propane–airexplosions using a baffle with a blockage ratio 30% (shown inAppendix A 2.4), during this experiment flame speeds reached closeto 1000 ms�1 and rates of pressure rise characteristic of detonativeburning were observed.

3.3. Ethylene experiments

Flame speeds recorded during selected ethylene–air explosionexperiments are shown in Fig. 7. Although flame speeds neverreached the predicted C–J detonation velocity of 1822 ms�1, it wasbelieved that detonative or detonative-like combustion took placeas pressure-time traces displayed the classic shape associated withdetonative combustion and rates of pressure rise of several

Table 2Flame speed and overpressure characteristics of propane explosions.

Configuration Max. flamespeed/ms�1

Max.overpressure(Smoothed)/bar

Max.overpressure(Peak)/bar

Max. rateof pressurerise/bar s�1

Average SD Average SD Average SD Average SD

Straight 72.3 13.1 1.7 0.6 2.2 1.1 5.9 1.6Bend 131.5 44.7 2.0 0.5 2.9 0.9 8.8 4.420% BR Baffle 133.1 43.9 1.8 0.2 2.5 0.4 7.1 2.1

Fig. 7. Flame speeds for ethylene–air explosions.

Table 3Run-up distances to DDT for hydrogen–air explosions.

Configuration Run-up distance to DDT

Average SD

Straight 12.1 0.7Bend 11.8 0.420% BR Baffle 10.9 0.5

R. Blanchard et al. / Journal of Loss Prevention in the Process Industries 23 (2010) 253–259258

thousand (bar s�1) were observed. After the DDT (at between 70and 80% of the pipe length) flame speeds decreased during thetransition from overdriven detonation to stable-detonation. Inter-actions with reflected pressure waves from the closed end of thepipe may also have acted to slow the flame/pressure front.The maximum flame speed for the straight pipe configuration wasobserved at approximately 80% of the pipe length, whereas for thetwo configurations containing obstacles this maximum was atapproximately 70%. On closer inspection of the flame speed profilesit can be seen that at approximately 20% of the pipe length(the obstacle is positioned at approximately 10% of the pipe length)the two obstacle containing configurations produced higher flamespeeds than the straight configuration. While these flames decel-erated slightly and returned to speeds closer to those whichwere observed in the straight configuration (at around 30% of thepipe length) they remained faster and therefore, assuming themaximum flame speed was associated with an overdriven deto-nation, DDT occurred over a shorter distance.

3.4. Hydrogen experiments

Flame speeds measured during three hydrogen–air explosionsusing the aforementioned pipe configurations are shown in Fig. 8.Each configuration used during these experiments produced

Fig. 8. Flame speeds for hydrogen–air explosions.

a transition from deflagration to denotation. Interpretation of theflame speed data showed that the point at which the transition fromfast deflagration to detonation occurred was between 60 and 80% ofthe pipe length, corresponding to between 68 and 96 pipe diame-ters. This transition was characterised by flame speeds over the CJdetonation velocity (1968 ms�1) which typically occur during DDTand was confirmed by analysis of the form of pressure-time signals.

Both the baffle and the bend had an initial effect on the flamespeed giving an acceleration after the obstacle. The largest increasein the flame speed, compared to the straight configuration, wasobserved with the 20% BR baffle where speeds of approximately800 ms�1 are observed immediately downstream of the obstacle. Theflame did not immediately transition into a detonation at this point,instead it returned to speeds similar to those seen in the straight pipeexperiment before the transition into a detonation occurred furtherdown the pipe (at approximately 60% of the pipe length). The sameeffects were observed for the 90 degree bend, however, the initialacceleration downstream of the obstacle was not as pronounced.

The run-up distance to DDTcan be determined by analysis of shockwaves from the overdriven detonation (Li, Lai, Chung, & Lu, 2005).A similar analysis was carried out for all hydrogen–air explosions,however, it was not possible for all ethylene and propane DDTs as theretonation shock waves from overdriven detonations were not strongenough to enable confident measurement of it’s velocity. The run-updistances to detonation for the hydrogen–air explosions described inthis section are shown in Table 3. While the bend and baffle config-urations showed similar flame speed profiles the baffle configurationwas able to affect DDT over a shorter distance than the bend config-uration. However, both bend and baffle configurations affected DDTover a shorter distance than the straight pipe configuration.

The flame acceleration caused by the obstacles was due toincreases in downstream turbulence leading to faster burning anda faster transition to detonation, however, at approximately 30%and 50% of the length of the pipe, for the baffle and bend obstaclesrespectively, the flame speeds returned to those seen with thestraight pipe configuration, these flames were then able to accel-erate at rates faster than flames in the straight pipe configuration toundergo DDT over a shorter distance. This could be due to effectscaused further downstream in the pipe after the initial accelerationor due to changes in the flame front shape at different points in thepipe. However in order to verify this hypothesis techniques such ashigh speed Schlieren or shadow photography would need to beemployed in order to better position and visualise the flame.

4. Conclusions

These experiments showed for the first time the ability of a full-bore obstacle to accelerate the burning velocity of a number ofgases and reduce the run-up distance required for DDT.

It was shown that a 90 degree bend placed at a relatively shortdistance from the ignition point in a long tube had the ability toenhance flame speeds and overpressures, and shorten the run-updistance to DDT. In terms of the qualitative effects on these param-eters they were comparable to a baffle type obstacle with a blockageratio of between 10 and 20%. It is expected that the flame speedenhancement caused by the 90 degree bend will be greater when the

R. Blanchard et al. / Journal of Loss Prevention in the Process Industries 23 (2010) 253–259 259

obstacle is placed further downstream of the ignition point, due tothe incoming flame having longer to accelerate and hence createa greater amount of turbulence downstream of the bend.

This work has contributed further to the argument that bends ina pipework system can have a significant effect on the combustionprocess, should be taken into account as part of a safety analysisand considered when placing explosion protection devices such asflame arresters or venting devices. Although this work has shownsome interesting effects there are still many questions unansweredregarding the effects of obstacle position, multiple bends and otherfull-bore obstacles.

Interesting observations were also made on the variation ofburning rates when using the same pre-set variables, an effect whichneeds to be further studied in order to determine the exact nature ofthe interactions which can contribute to flame acceleration.

Appendix. Supplementary information

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.jlp.2009.09.004

References

Andrews, G. E., Herath, P., & Phylaktou, H. N. (1990). The influence of flow blockageon the rate of pressure rise in large L/D cylindrical closed vessel explosions.Journal of Loss Prevention in the Process Industries, 3(3), 291–302.

Berger, S. A., Talbot, L., & Yao, L. S. (1983). Flow in curved pipes. Annual Review ofFluid Mechanics, 15, 461–512.

Burgess, M. J. (1971). Pressures losses in ducted flows. London: Butterworth.Chatrathi, K. (1992). Deflagration protection of pipes. Plant/Operations Progress,

11(2), 116–120.Ciccarelli, G., & Dorofeev, S. (2008). Flame acceleration and transition to detonation

in ducts. Progress in Energy and Combustion Science, 34(4), 499–550.Clanet, C., & Searby, G. (1996). On the ‘tulip flame’ phenomenon. Combustion and

Flame, 105(1–2), 225–238.European Standard EN 12874. (2001). Flame arresters – Performance requirements,

test methods and limits for use.Frolov, S. M. (2008). Fast deflagration-to-detonation transition. Russian Journal of

Physical Chemistry B, 2(3), 442–455.Gonzalez, M., Borghi, R., & Saouab, A. (1992). Interaction of a flame front with its

self-generated flow in an enclosure – the tulip flame phenomenon. Combustionand Flame, 88(2), 201–220.

Green, D. W. (1999). Perry’s chemical engineers’ handbook. New York: McGraw-Hill.Grossel, S. S. (2002). Deflagration and detonation flame arresters. New York: Amer-

ican Institute of Chemical Engineers.Ibrahim, S. S., & Masri, A. R. (2001). The effects of obstructions on overpressure

resulting from premixed flame deflagration. Journal of Loss Prevention in theProcess Industries, 14(3), 213–221.

Li, J., Lai, W. H., Chung, K., & Lu, R. K. (2005). Uncertainty analysis of deflagration-to-detonation run-up distance. Shock Waves, 14(5–6), 413–420.

Lohrer, C., Drame, C., Schalau, B., & Gratz, R. (2008). Propane/air deflagrations andCTA measurements of turbulence inducing elements in closed pipes. Journal ofLoss Prevention in the Process Industries, 21(1), 1–10.

Lohrer, C., Hahn, M., Arndt, D., & Gratz, R. (2008). Einfluss eines 90�-Rohrbogens ineiner technischen Rohrleitung auf reaktive Stromungen. (in German). ChemieIngenieur Technik, 80(5), 649–657.

Masri, A. R., Ibrahim, S. S., Nehzat, N., & Green, A. R. (2000). Experimental study ofpremixed flame propagation over various solid obstructions. ExperimentalThermal and Fluid Science, 21(1–3), 109–116.

Newsholm, G. (2004). Safe design and use of vent collection systems for potentiallyflammable mixtures. HSE Online.

Oakley, G. I., & Thomas, G. O. (2004). Investigations into concerns about BS EN12874:2001 flame arresters. Performance requirements, tests methods andlimits for use. HSE Research Report 281.

Phylaktou, H., & Andrews, G. E. (1991a). Gas explosions in long closed vessels.Combustion Science and Technology, 77(1–3), 27–39.

Phylaktou, H., & Andrews, G. E. (1991b). The acceleration of flame propagation ina tube by an obstacle. Combustion and Flame, 85(3–4), 363–379.

Phylaktou, H., Foley, M., & Andrews, G. E. (1993). Explosion enhancement througha 90-degress curved bend. Journal of Loss Prevention in the Process Industries,6(1), 21–29.

Roy, G. D., Frolov, S. M., Borisov, A. A., & Netzer, D. W. (2004). Pulse detonationpropulsion: challenges, current status, and future perspective. Progress inEnergy and Combustion Science, 30(6), 545–672.

Sato, K., Sakai, Y., & Chiga, M. (1996). Flame propagation along 90� bend in an openduct. In Twenty-sixth symposium on combustion (pp. 931–937). The CombustionInstitute.

Starke, R., & Roth, P. (1989). An experimental investigation of flame behavior duringexplosions in cylindrical enclosures with obstacles. Combustion and Flame,75(2), 111–121.

Zhou, B., Sobiesiak, A., & Quan, P. (2006). Flame behavior and flame-induced flow ina closed rectangular duct with a 90 degrees bend. International Journal ofThermal Sciences, 45(5), 457–474.

Nomenclature

BR: Blockage ratioCTA: Constant temperature anemometryDDT: Detonation to deflagration transitionSD: Standard deviationVOC: Volatile organic solvent